[hafqa] [hafqa] [Bug 3150] segfault in libfontconfig when using -O0 in application / nokia sans font

[From: bugzilla-daemon at maemo.org]

https://bugs.maemo.org/show_bug.cgi?id=3150





------- Comment #8 from eero.tamminen at nokia.com  2008-10-13 13:21 GMT+3 -------
> Backtrace to be attached.

Thanks!

> I used the default libraries at this time with the default optimizations,
> whatever they are.  I had to add the diablo/tools and diablo/sdk repositories
> from repository.maemo.org to install gdb and the debug symbols.

This seems very strange.

The crash is here (I sanitized the bt a bit):
#0  FcPatternObjectAddWithBinding () at fcpat.c:475
#1  0x40e21494 in FcPatternObjectAdd () at fcpat.c:544
#2  0x40e215e6 in FcPatternObjectAddString () at fcpat.c:663
#3  0x40e21610 in IA__FcPatternAddString () at fcpat.c:669
#4  0x40e9f3b0 in pango_fc_font_map_load_fontset () at pangofc-fontmap.c:852

And the only place where the backtrace goes through -O0 compiled application is
here, 61 stack frames up):
#60 0x40651ed4 in _gtk_size_group_compute_requisition
(widget=0xc2890,requisition=0xbef8b2dc) at gtksizegroup.c:817
#61 0x000171c4 in iter (widget=0xc2890, data=0x0) at main.c:2444
#62 0x406b1ddc in gtk_toolbar_forall () at gtkcontainer.c:1563
#64 0x00019328 in main (argc=1, argv=0xbef8b6c4) at main.c:2448

I don't see how application compile options could affect font handling unless
the application is corrupting its memory somehow and which then only on ARM
triggers a crash...

If you Valgrind the unoptimized application on x86, does it report any memory
access errors?

If you build the application for ARM using "-Wall -O2", what kind of
compilation warnings the compiler gives (code errors can manifest in different
ways with different optimization levels)?


-- 
Configure bugmail: https://bugs.maemo.org/userprefs.cgi?tab=email
Replies to this email are NOT read, instead please add comments at
https://bugs.maemo.org/show_bug.cgi?id=3150
------- You are receiving this mail because: -------
You are the QA contact for the bug, or are watching the QA contact.